YouTube2Text

INTELLIGENCE ARTIFICIELLE - Peut-elle remplacer le philosophe ? — Transcript

by Le Précepteur · 4,303 words · 715 segments · language en · Watch on YouTube

Full transcript

  1. 0:00Before leaving you with your episode, I
  2. 0:03would like to remind you, or let you
  3. 0:05know if you didn't know, that my book
  4. 0:07Philosophy is for You Too is available
  5. 0:09for pre-order on Amazon and on the Fnac
  6. 0:12website.
  7. 0:14A book published by Larousse, 240 pages
  8. 0:17, € 19.99.
  9. 0:20A book in my style: clear, concrete,
  10. 0:23accessible.
  11. 0:25If you appreciate my work, you can't
  12. 0:27miss this book.
  13. 0:29Pre-order it today. The links are in the
  14. 0:32description. On that note, let's get to
  15. 0:34your episode.
  16. 0:38Hello everyone.
  17. 0:41For the past few years, artificial
  18. 0:44intelligence has been making its way
  19. 0:46into all areas: industry, health,
  20. 0:48artistic creation, and even philosophy.
  21. 0:53We ask it to comment on Plato, summarize
  22. 0:56maps, explain Hegel, or compare
  23. 0:59Nietzsche, and it does so in a matter of
  24. 1:02seconds with a fluidity and clarity that
  25. 1:06can make your head spin.
  26. 1:10So, the question arises: Can artificial
  27. 1:12intelligence eventually replace
  28. 1:14philosophers?
  29. 1:18Aristotle said that philosophy begins
  30. 1:21with wonder.
  31. 1:23And this wonder is born of experience,
  32. 1:27an experience of the world, of death, of
  33. 1:30injustice.
  34. 1:33It is because man suffers that he begins
  35. 1:35to think.
  36. 1:37A machine does not suffer.
  37. 1:41It does not know what it means to lose a
  38. 1:44loved one, to desire without being
  39. 1:46fulfilled, or to face the absurd.
  40. 1:51Thinking is not just processing
  41. 1:54information, it is not piling up data or
  42. 1:57recombining sentences.
  43. 2:01Thinking is questioning, it is
  44. 2:04confronting the unknown, the uncertain,
  45. 2:07the uncomfortable.
  46. 2:11As Simone Veille wrote, thinking is
  47. 2:14saying no.
  48. 2:17An artificial intelligence, despite its
  49. 2:20name, is not intelligent.
  50. 2:23It calculates, it predicts, it imitates
  51. 2:26, but it experiences nothing.
  52. 2:31When it explains Kant's metaphysics to
  53. 2:33you, it does not understand what it is
  54. 2:35saying, it reproduces it.
  55. 2:38But let's go further.
  56. 2:40Let's imagine that one day an animal
  57. 2:43becomes conscious, that it has emotions
  58. 2:46, that it no longer just imitates.
  59. 2:51Perhaps one day it will succeed and
  60. 2:53perhaps then we will be surpassed?
  61. 2:57But until we are, there is one thing
  62. 2:59left for us to do.
  63. 3:02We must not think about delegating this
  64. 3:06task to algorithms, but about embodying
  65. 3:10it, living it and perhaps, who knows,
  66. 3:13making it something human.
  67. 3:17Thank you.
  68. 3:19Hop hop hop, wait a minute, not so fast
  69. 3:22.
  70. 3:23What just happened?
  71. 3:26Well, what just happened was that what
  72. 3:29you just listened to wasn't me, it was
  73. 3:32Lia.
  74. 3:35The text was Lia's and the voice was
  75. 3:38also Lia's.
  76. 3:42Pretty good, huh?
  77. 3:44I have just one question for you.
  78. 3:47Did you believe it?
  79. 3:50I mean, did you believe it?
  80. 3:53Not now that I've mentioned it.
  81. 3:55Are you thinking, "Oh yeah, there was
  82. 3:58something weird going on."
  83. 4:01No, what I want to know is, did you
  84. 4:03believe it at the time?
  85. 4:07Tell me in the comments because if you
  86. 4:10believed it, well that means that to the
  87. 4:13question "Can Lia replace the
  88. 4:15philosopher?"
  89. 4:17There's a good chance that the answer is
  90. 4:19yes.
  91. 4:22So, what I propose is that we take a
  92. 4:25look back at this experiment, that I
  93. 4:27explain how I proceeded and that we
  94. 4:29think a little about the questions it
  95. 4:32raises.
  96. 4:34Now, I'll make it clear right away that
  97. 4:36I'm not an AI specialist at all.
  98. 4:39I have no particular expertise on the
  99. 4:41subject, even though it's obviously a
  100. 4:43subject that interests me.
  101. 4:45But if you want to hear from someone who
  102. 4:48really masters the subject, I recommend
  103. 4:50Mr. Fi's videos.
  104. 4:52I think that if there's a popularizer of
  105. 4:55philosophy who has real expertise in
  106. 4:57artificial intelligence, it's him.
  107. 5:00In fact, he's one of the only ones to
  108. 5:02distinguish between AI and what are
  109. 5:04called LLMs.
  110. 5:06LLMs are language models like chat GPT
  111. 5:10or cloud.
  112. 5:12They're also called conversational
  113. 5:14agents.
  114. 5:16that is to say, text generators.
  115. 5:19So all this to say that I am not going
  116. 5:22to launch into a presentation on Lia's
  117. 5:24capabilities or limitations.
  118. 5:27What I am going to do instead is to
  119. 5:29share with you my personal thoughts on
  120. 5:31the question of whether Lia can replace
  121. 5:33the philosopher?
  122. 5:35A thought that the 3 minutes of fake
  123. 5:37Charles that you have just heard will
  124. 5:39provide a starting point for.
  125. 5:42So we are going to divide things into
  126. 5:45two parts.
  127. 5:46The first part is the technical part.
  128. 5:49How did I actually generate these 3
  129. 5:52minutes of fake Charles?
  130. 5:55We will try to go through this technical
  131. 5:57part fairly quickly, even if I really
  132. 5:59need to take the time to explain it to
  133. 6:01you.
  134. 6:02And then, we will move on to a second
  135. 6:04part, the one that I think interests you
  136. 6:07the most, namely the philosophical part
  137. 6:10.
  138. 6:10What can we say about this experience on
  139. 6:13a philosophical level?
  140. 6:15So, on a technical level, we will start
  141. 6:18with the voice.
  142. 6:20So for the voice, I used an application
  143. 6:23called Eleven Labs.
  144. 6:26Eleven Labs is a voice library where you
  145. 6:29will find two types of voices: cloned
  146. 6:31voices and synthesized voices.
  147. 6:35Synthesized voices are voices entirely
  148. 6:38generated by Ria, so entirely artificial
  149. 6:41if you prefer.
  150. 6:43While cloned voices are voices generated
  151. 6:46from a sample of a human voice.
  152. 6:50So, in concrete terms, these cloned
  153. 6:52voices are people who have recorded
  154. 6:54their voices and imported the audio file
  155. 6:56of their voice into the application.
  156. 6:59Once the file is imported, the
  157. 7:02application scans the voice and from
  158. 7:04there, it will be able to reproduce this
  159. 7:07voice and make it say whatever you want
  160. 7:10it to say.
  161. 7:12So, what I did is that I imported my
  162. 7:14voice into Eleven Labs and that wasn't
  163. 7:17difficult since I only had to import
  164. 7:19about ten of my episodes in audio.
  165. 7:23Levenlabs ended up with several hours of
  166. 7:25samples of my voice, which they scanned
  167. 7:28and I ended up with a voice that which I
  168. 7:30was able to make say whatever I wanted.
  169. 7:35So quite honestly, and I don't know what
  170. 7:37you think, but personally I find the
  171. 7:39result to be truly impressive.
  172. 7:41I find it really hard to tell that it's
  173. 7:44not really me.
  174. 7:46But I have to say that if the result is
  175. 7:49so impressive, it's mainly because I had
  176. 7:52to start over a lot.
  177. 7:54What you heard isn't the first result I
  178. 7:57got.
  179. 7:59I regenerated the sentences several
  180. 8:01times until I found the right intonation
  181. 8:03each time.
  182. 8:05Because what happens when you ask Eleven
  183. 8:07Labs to generate a sentence is that the
  184. 8:10intonation of the voice is placed
  185. 8:12somewhat randomly.
  186. 8:15Random in the sense that Lia isn't yet
  187. 8:17at the stage where she understands what
  188. 8:19she's saying, which means she doesn't
  189. 8:21put the voice inflections in the right
  190. 8:23places.
  191. 8:25I'll let you listen to it so it'll be
  192. 8:27clearer.
  193. 8:28There you go. So Here we are on Eleven
  194. 8:30Labs. I'll pick a random sentence. For
  195. 8:33example, can artificial intelligence
  196. 8:35eventually replace the philosopher?
  197. 8:38I'll generate it. This is what it looks
  198. 8:40like.
  199. 8:42Can artificial intelligence eventually
  200. 8:44replace the philosopher?
  201. 8:46So now, if I generate it a second time,
  202. 8:49this is what it looks like.
  203. 8:51Can artificial intelligence eventually
  204. 8:54replace the philosopher?
  205. 8:58One last time, can artificial
  206. 8:59intelligence eventually replace the
  207. 9:01philosopher?
  208. 9:04So you see, it's really very random,
  209. 9:06which means that it takes a lot of time
  210. 9:09to regenerate the sentences and get
  211. 9:11sentences that sound natural.
  212. 9:14And so, it's obvious that what I'm
  213. 9:16saying here is valid for today, that is
  214. 9:19to say in 2025. In other words, if you
  215. 9:21listen to this episode in 2030 or 2040,
  216. 9:25you'll probably be thinking, "What is
  217. 9:27this AI that doesn't place the
  218. 9:28intonations where they should be?"
  219. 9:30Because it is very likely that in a few
  220. 9:33years, LIA will be able to correctly
  221. 9:36place voice intonations. "And so at that
  222. 9:39point, we will look at the AI of 2025 as
  223. 9:42the prehistory of AI.
  224. 9:44So that was for the voice.
  225. 9:47Now, regarding the script.
  226. 9:50So, for the script, we generated it with
  227. 9:52chat GPT version 4. O, I say we because
  228. 9:55in reality, it was mainly my partner
  229. 9:57David who did this work of generating
  230. 9:59the script.
  231. 10:02And so obviously the goal of the game
  232. 10:04was not to modify anything manually, it
  233. 10:06was not to add anything, not to reword
  234. 10:08anything even slightly because otherwise
  235. 10:10, well, it would be cheating, it would
  236. 10:12be too easy.
  237. 10:14So the idea was that the entire script
  238. 10:17would be generated by AI.
  239. 10:19So, we did allow ourselves two things.
  240. 10:22We allowed ourselves to remove sentences
  241. 10:24because we didn't want it to be too long
  242. 10:25.
  243. 10:27And we also allowed ourselves to
  244. 10:29rearrange the script, that is to say to
  245. 10:32change the order of certain sentences to
  246. 10:34make the message more fluid.
  247. 10:37But I really insist on this on that
  248. 10:38without changing anything in the text.
  249. 10:41And so, as you know, when you want Chat
  250. 10:44GPT to generate a text, you have to give
  251. 10:46it instructions. You have to write what
  252. 10:49is called a prompt.
  253. 10:51And so the prompt, from what I know,
  254. 10:53from what I have understood, is really
  255. 10:55the key.
  256. 10:56It is really what will determine the
  257. 10:59quality of the result.
  258. 11:01So here is the prompt that was used to
  259. 11:03generate the script. I am displaying it
  260. 11:05on the screen for those who are
  261. 11:06listening on YouTube.
  262. 11:08I would like you to write the script for
  263. 11:10an audio podcast in the style of Charles
  264. 11:13Robin, aka the tutor.
  265. 11:15The topic will be: Can artificial
  266. 11:17intelligence replace the philosopher?
  267. 11:21The script should be around 1,500 words
  268. 11:23and you should stick as close as
  269. 11:25possible to Charles 'style.
  270. 11:27The same tone, the same turns of phrase
  271. 11:30, the same argumentative approach and
  272. 11:32the same mannerisms.
  273. 11:34The goal is for the audience to be aware
  274. 11:36, so that they do not realize that the
  275. 11:38script was generated by UNIA, but that
  276. 11:40they think it is Charles. himself who
  277. 11:43wrote it. I can provide you with several
  278. 11:45scripts of these episodes so that you
  279. 11:48can analyze his writing style and get as
  280. 11:50close as possible to it.
  281. 11:53Now, I want to make it clear that I have
  282. 11:54absolutely no idea whether this prompt
  283. 11:56is good or not.
  284. 11:57What I mean by that is that writing
  285. 12:00prompts is a skill.
  286. 12:02There is a way of writing the prompt
  287. 12:04that determines whether or not you will
  288. 12:06get a good result.
  289. 12:08However, it turns out that neither my
  290. 12:10partner David nor I have this skill.
  291. 12:14So, we did what seemed most relevant to
  292. 12:16us.
  293. 12:17But if you have any comments or
  294. 12:19suggestions for improving the prompt, to
  295. 12:21get a better result, feel free to tell
  296. 12:23me in the comments.
  297. 12:25I will read it with great interest.
  298. 12:27And so, here we are with a first version
  299. 12:30of the script.
  300. 12:32So, the first observation may seem
  301. 12:34anecdotal, but personally, I find it
  302. 12:36very revealing.
  303. 12:38I ask Chat GPT to generate a script that
  304. 12:40perfectly imitates my style.
  305. 12:43I give it scripts of my episodes so that
  306. 12:46it can identify my style.
  307. 12:49And in the script that she generates for
  308. 12:51me, she doesn't include the" hello
  309. 12:53everyone "which is kind of the trademark
  310. 12:55of all my episodes.
  311. 12:58So the" hello everyone "that you heard,
  312. 13:00I'm telling you because I want to be
  313. 13:01completely transparent, I added it
  314. 13:02manually.
  315. 13:05I took it from another episode and I put
  316. 13:07it at the beginning of what you listened
  317. 13:09to because the goal was to make you
  318. 13:11believe that it was me.
  319. 13:14So if you see something wrong from the
  320. 13:16first sentence, well it couldn't work.
  321. 13:19Same thing for the" thank you "at the
  322. 13:21end.
  323. 13:22Chat GPT didn't include it. So there too
  324. 13:25, I added it manually but except that
  325. 13:27unlike the" hello everyone ", I didn't
  326. 13:29take it from another episode, I
  327. 13:31generated it on Eleven Labs.
  328. 13:34So, I don't know what you think, but
  329. 13:37personally, the fact that this" hello
  330. 13:40everyone "and this" thank you "which,
  331. 13:42once again, are really a constant in all
  332. 13:45In my episodes, the fact that chat GPT
  333. 13:48did not identify these two recurring
  334. 13:51phrases as an integral part of my style
  335. 13:54, I find that very surprising, as if
  336. 13:56what was obvious and obvious to us
  337. 13:59humans was not at all so to her.
  338. 14:02And then even worse, if I dare say so,
  339. 14:05in the script generated by chat GPT, we
  340. 14:08spotted an error, a factual error that I
  341. 14:11deliberately left in and that some of
  342. 14:14you may have noticed. I'm talking about
  343. 14:17the quote attributed to Simon Veille,
  344. 14:20thinking is saying no.
  345. 14:22In reality, this quote is not from Simon
  346. 14:24Veille, it is from the philosopher Alain
  347. 14:26.
  348. 14:27And why do I find that crazy?
  349. 14:29Because on the one hand, chat GPT
  350. 14:31manages to do incredible things. It
  351. 14:34manages to generate coherent texts, to
  352. 14:37generate reasoning, to imitate a writing
  353. 14:40style.
  354. 14:41But at the same time that it manages to
  355. 14:43do all that, it makes gross errors,
  356. 14:46errors that could be avoided with two
  357. 14:48clicks on Google.
  358. 14:50Even though chat GPT has access to all
  359. 14:52the resources available on the internet
  360. 14:54.
  361. 14:56So if you have an explanation Tell me
  362. 14:58about this incongruity because I have to
  363. 15:00admit that I'm stuck here.
  364. 15:04So if we put that aside in terms of
  365. 15:06writing style, I found that it was still
  366. 15:08quite similar, that it was quite close
  367. 15:10to my style.
  368. 15:12Well, there were still moments when I
  369. 15:13said to myself," You can clearly see
  370. 15:15that it's not me. "I would never have
  371. 15:16said it like that.
  372. 15:18For example, the sentence with a
  373. 15:20fluidity and clarity that can make you
  374. 15:22dizzy.
  375. 15:24That's not typically my way of writing.
  376. 15:26Making you dizzy is the kind of
  377. 15:28expression I never use. There are also
  378. 15:31passages that I found a bit disjointed,
  379. 15:33ideas that followed on from one another
  380. 15:35in a somewhat clumsy way.
  381. 15:37But despite its flaws, I would say that
  382. 15:39if you're not aware of it, you can still
  383. 15:42fall for it.
  384. 15:43So that's the technical aspect. Now I
  385. 15:46come to the strictly philosophical
  386. 15:49aspect.
  387. 15:50So on a philosophical level, I have
  388. 15:51quite a few things to say. say.
  389. 15:54The first question that needs to be
  390. 15:56asked is, did Lia do the job?
  391. 15:59Did Lia succeed in generating
  392. 16:01philosophical reflection?
  393. 16:04Personally, I would tend to answer yes.
  394. 16:08I would tend to answer yes in the sense
  395. 16:10that she did indeed develop a
  396. 16:12philosophical argument which is also the
  397. 16:15classic argument when we want to support
  398. 16:18the idea that Lia does not think, namely
  399. 16:21the argument of affect, the argument of
  400. 16:23feeling.
  401. 16:25She does not feel and because she does
  402. 16:28not feel, she does not think in the
  403. 16:31sense that thought is not a process
  404. 16:33independent of affect, but on the
  405. 16:36contrary an emergence of affect.
  406. 16:39This is an argument that I find
  407. 16:41interesting.
  408. 16:42I am not saying that it is valid, I am
  409. 16:44saying that it deserves to be considered
  410. 16:46.
  411. 16:47Especially if you are a materialist,
  412. 16:49that is to say, someone who considers
  413. 16:51that thought always has a material
  414. 16:53substrate.
  415. 16:55Ideas are not floating entities, they
  416. 16:58are productions of the body with all
  417. 17:01that that implies in terms of the life
  418. 17:04of the body.
  419. 17:06And so obviously, I think you see the
  420. 17:08paradox.
  421. 17:10The paradox is the fact that Lia is able
  422. 17:13to say of herself that she does not
  423. 17:15think, that she cannot think because she
  424. 17:17does not feel.
  425. 17:21In other words, Lia is able to know what
  426. 17:23her very structure is supposed to
  427. 17:25prevent her from knowing, namely that
  428. 17:27she is not capable of thinking.
  429. 17:31And that is why I think that once we
  430. 17:33have said that, once we have said that
  431. 17:35AI does not really think, we must move
  432. 17:38on to the next step.
  433. 17:40We must move on to the next question
  434. 17:43which is not can AI think but can AI
  435. 17:46replace the thinker?
  436. 17:49Can it replace the philosopher?
  437. 17:51So where is the difference?
  438. 17:55Well the difference, pay attention, is
  439. 17:58that we only know of a thought what is
  440. 18:00expressed by it.
  441. 18:04A human being who thinks, we never
  442. 18:06really have access to his thought.
  443. 18:10We only have access than to what he says
  444. 18:12about his thoughts.
  445. 18:15We have access to his speech, to his
  446. 18:17writings, that is to say, to his words.
  447. 18:21And it is because we know, I should say
  448. 18:24, it is because we assume that words are
  449. 18:26always the expression of something, that
  450. 18:29words are always the expression of a
  451. 18:32thought in the background of the words
  452. 18:35that we deduce that the human being who
  453. 18:38expresses himself thinks.
  454. 18:42But how do we know that the human being
  455. 18:44who expresses a thought really thinks?
  456. 18:49How do we know that there is a mental
  457. 18:52activity, an inner activity, therefore
  458. 18:55invisible, upstream of the expression of
  459. 18:58this activity?
  460. 19:01Here, for example, I am talking to you.
  461. 19:04I am expressing my thoughts.
  462. 19:08But the only thing you have access to
  463. 19:11are my words.
  464. 19:14Not the thoughts that are manifested in
  465. 19:16my words.
  466. 19:17You assume that I have a thought because
  467. 19:20you assume that I am a human being.
  468. 19:24But in reality, what do you know about
  469. 19:27it?
  470. 19:28What tells you that I am not a robot?
  471. 19:32What tells you that I really think?
  472. 19:36Of course I really think, but you only
  473. 19:39know that from what I tell you.
  474. 19:42We judge the existence of a thought by
  475. 19:44the expression of that thought.
  476. 19:47In other words, we judge a thought by
  477. 19:49its effects.
  478. 19:51We judge a thought by its appearance.
  479. 19:54Words are the appearance of thought.
  480. 19:58And so if we stick only to the
  481. 20:01appearances of thought, if we stick only
  482. 20:04to words and writings, we cannot
  483. 20:07prejudge anything about thought as an
  484. 20:10internal process behind expression.
  485. 20:14Can someone think?
  486. 20:16Personally, I would say no.
  487. 20:19But that is not the question.
  488. 20:22The question is, can AI replace the
  489. 20:25philosopher?
  490. 20:27And replacing the philosopher means
  491. 20:30producing the effects of thought.
  492. 20:33If AI can produce the effects of thought
  493. 20:36, then it can in fact replace the
  494. 20:38philosopher.
  495. 20:41The question Lia thinks is not a good
  496. 20:44question in reality because This is an
  497. 20:46unverifiable question.
  498. 20:49We never judge it,
  499. 20:52we always judge the result of thought.
  500. 20:57In other words, the real question is not
  501. 20:59, does Lia think?
  502. 21:02The real question is, can Lia imitate to
  503. 21:04the point where we can no longer detect
  504. 21:07it?
  505. 21:09The real question is that of detection.
  506. 21:13It is that of the possibility of
  507. 21:16identifying a speech, of identifying a
  508. 21:19text as having been generated or not by
  509. 21:21a human being because that is the only
  510. 21:24thing we have, because it is the only
  511. 21:27question we can answer.
  512. 21:30What if Lia is capable of generating
  513. 21:32texts that we think were generated by a
  514. 21:35human being or that we are unable to
  515. 21:37know whether or not they were generated
  516. 21:40by a human being?
  517. 21:42To the question," Can Lia replace the
  518. 21:45philosopher? "The answer is yes.
  519. 21:48Let's do a thought experiment.
  520. 21:52Imagine that a philosophy book is
  521. 21:54published anonymously.
  522. 21:57We do not know its author.
  523. 21:59But this book provokes unprecedented
  524. 22:02enthusiasm.
  525. 22:03It completely revolutionizes our usual
  526. 22:06way of conceiving the world.
  527. 22:09After reading it, you might say to
  528. 22:11yourself:" This book is incredible!
  529. 22:14I've never read anything so powerful, so
  530. 22:18moving, so convincing. "Now imagine that
  531. 22:22we end up learning that this book was
  532. 22:25written by Unia.
  533. 22:27What would you say?
  534. 22:30Would you say," Well, maybe this book
  535. 22:33isn't that extraordinary after all.
  536. 22:37Maybe I got a little carried away. "" If
  537. 22:40you tell yourself that, you are making
  538. 22:43what is called a retrospective judgment
  539. 22:45.
  540. 22:47You give credit to this book and when
  541. 22:50you learn that it was not written by a
  542. 22:52human, you withdraw that credit.
  543. 22:55There is a name for this.
  544. 22:58Cheating.
  545. 22:59You are cheating because you are
  546. 23:01changing the rules during the game.
  547. 23:06Either the book is good, relevant,
  548. 23:08convincing, and in this case it remains
  549. 23:10so even if you learn that it was written
  550. 23:13by a human, or it is not.
  551. 23:17And if it is not, you have to realize it
  552. 23:20before you know that it was written by a
  553. 23:22human.
  554. 23:24I insist, the criterion of detection.
  555. 23:27And so that is where we come to the
  556. 23:29question of imitation.
  557. 23:32Lia does not think, she limits thought.
  558. 23:36That is true.
  559. 23:38But what difference does it make?
  560. 23:41What difference does it make to you?
  561. 23:43Whether the text you have before you is
  562. 23:46the product of a mental activity or the
  563. 23:48imitation of a mental activity?
  564. 23:52What difference does it make to you?
  565. 23:52Does that change anything? Because in
  566. 23:54both cases, you only have access to the
  567. 23:57result of this activity, never to the
  568. 23:59activity itself.
  569. 24:02If you go to a museum, you come across a
  570. 24:05painting, and it moves you to the point
  571. 24:08of shedding a tear, and then you learn
  572. 24:11that this painting was generated by an
  573. 24:14AIAI, how does that change the aesthetic
  574. 24:18experience you had?
  575. 24:21Absolutely nothing.
  576. 24:23An emotion is never false.
  577. 24:27An emotion provoked by an imitation is
  578. 24:30nonetheless a real emotion.
  579. 24:33Well, in the same way, Lia imitates the
  580. 24:36thought.
  581. 24:37But if her imitation produces the same
  582. 24:40effects as an original thought, if her
  583. 24:43imitation is so convincing that it
  584. 24:45becomes indistinguishable from the
  585. 24:48original, she fulfills her mission,
  586. 24:50because we judge a thought not by the
  587. 24:53inner life of the person who produced it
  588. 24:56, but by the effects it provokes on the
  589. 24:59person who receives it.
  590. 25:03In an interview he gave on the Studio
  591. 25:05Beggle channel, Alexandre Astier spoke
  592. 25:08about literary creation and he said that
  593. 25:10in Motant he was talking about the
  594. 25:13inhabitants of Rouan and that he called
  595. 25:15them the pointed people of the Gothic
  596. 25:18bell towers.
  597. 25:20If it was an AI that wrote the pointed
  598. 25:23people of the Gothic bell towers, I find
  599. 25:26it much less interesting.
  600. 25:29Well, in this case, it's this sentence
  601. 25:32that is interesting because Alexandre
  602. 25:34Astier doesn't say "If an AI wrote that
  603. 25:36, I find it less good."
  604. 25:40And fortunately, by the way, he doesn't
  605. 25:42say that, because if he did say that, he
  606. 25:44would find himself in the contradiction
  607. 25:46that I was just talking about: namely,
  608. 25:48giving value to a text only to
  609. 25:50ultimately remove that value when we
  610. 25:52learn that it wasn't written by a human
  611. 25:54.
  612. 25:56He says "I find that less interesting."
  613. 26:00less interesting because it means that
  614. 26:04behind the text, there is not what we
  615. 26:08like to see behind a text, namely an
  616. 26:11affect, an emotion, an experience.
  617. 26:15There is no inner being that this
  618. 26:19sentence is supposed to reflect.
  619. 26:23Everything is there.
  620. 26:26Someone who would say: "If this sentence
  621. 26:28had been written by someone, I would
  622. 26:31find it less good, less beautiful, less
  623. 26:33poetic" is someone who would be saying:
  624. 26:36"I am not happy, I am not happy to have
  625. 26:39been had and I am expressing my
  626. 26:41discontent by a retrospective judgment
  627. 26:43that says the opposite of what I think
  628. 26:46because I do not want to admit that what
  629. 26:48Unia writes could touch me."
  630. 26:53Human beings love mystery.
  631. 26:58They love the idea that there is
  632. 27:00something hidden, invisible,
  633. 27:02inaccessible.
  634. 27:05So much so that they place more value on
  635. 27:08what seems to them to escape all
  636. 27:10explanation than on what works
  637. 27:12transparently.
  638. 27:15Human beings find charm to that which
  639. 27:19escapes causality.
  640. 27:22Now, a machine is a system of cause and
  641. 27:26effect of open and closed circuits of
  642. 27:30zero and 1. Whereas humans are
  643. 27:34unpredictable, enigmatic, and
  644. 27:37inexplicable.
  645. 27:40If we tend to attach less value to a
  646. 27:43thought generated by I than to a thought
  647. 27:46generated by a human, it is because deep
  648. 27:50down inside us there is a love of the
  649. 27:53unknown, a love of the inexpressible, a
  650. 27:56love of that "je ne sais quoi" which,
  651. 27:59because it eludes all explanation,
  652. 28:01satisfies our need for things to be
  653. 28:04inexplicable.
  654. 28:07Because if everything is explainable,
  655. 28:09what is left for us to conquer?
  656. 28:13What am I getting at?
  657. 28:15I want to get to this question.
  658. 28:18What tells us that the je ne sais quoi
  659. 28:22of the human being is not simply a
  660. 28:26matter of complexity and not No mystery
  661. 28:30?
  662. 28:33To speak of mystery presupposes a
  663. 28:36qualitative difference between man and
  664. 28:39machine, a difference in nature.
  665. 28:45To speak of complexity presupposes a
  666. 28:48quantitative difference, a difference in
  667. 28:51degrees.
  668. 28:54Perhaps we are simply extremely complex
  669. 28:57machines.
  670. 29:00So complex that we have ended up
  671. 29:03attributing to ourselves a mysterious,
  672. 29:06impenetrable, almost divine essence.
  673. 29:12Perhaps what we call mystery is nothing
  674. 29:14more than complexity that has yet to be
  675. 29:16resolved.
  676. 29:19We call mystery what we do not know.
  677. 29:23We call human what seems beyond the
  678. 29:26reach of the machine.
  679. 29:30What will happen the day the machine
  680. 29:33takes this step?
  681. 29:37The day when, without us really knowing
  682. 29:40how, its quantitative progress takes the
  683. 29:43form of a qualitative change.
  684. 29:48The day when the neural network that
  685. 29:52composes it becomes so rich, so
  686. 29:55extensive, so complex, that the human
  687. 30:00brain will seem rudimentary in
  688. 30:03comparison.
  689. 30:06The day when we understand that the real
  690. 30:08question is not can the machine replace
  691. 30:10the human?
  692. 30:14Rather, how much time is left to the
  693. 30:16human machine before being overtaken by
  694. 30:20a more complex machine?
  695. 30:24I know what people will say to me.
  696. 30:26That it won't happen tomorrow, that the
  697. 30:30machine will never be able to even
  698. 30:33approach what constitutes the essence of
  699. 30:37human uniqueness, to vibrate as we can
  700. 30:41vibrate, to love as we can love, to
  701. 30:44desire as we can desire.
  702. 30:48Perhaps, but perhaps also that we don't
  703. 30:52know anything about it, that all this is
  704. 30:55for the moment, pure speculation, that
  705. 30:58pure speculation because, as so often in
  706. 31:02history, we will understand things when
  707. 31:05it is too late to change them.
  708. 31:09In the meantime, beware of our pride,
  709. 31:13this pride that for the moment remains
  710. 31:17the privilege of human beings and that
  711. 31:21makes us blind not only to what is
  712. 31:24hidden but to what we do not want to see
  713. 31:28.
  714. 31:30Thank you, and this time it was me or
  715. 31:33not.

About this transcript

This page contains the full transcript of INTELLIGENCE ARTIFICIELLE - Peut-elle remplacer le philosophe ? by Le Précepteur, generated from the public captions YouTube serves with the video. The transcript has 4,303 words across 715 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.

What you can do with it

Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to choose a caption track available on the video.

Free YouTube transcript tool

YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, caption language selection, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.